os/bluestore: refactor dirty blob tracking along with some related fixes #10215
Conversation
ifed01
changed the title
ifed01
force-pushed
the
wip-bluestore-blob-release-fix
branch
from
2d29f9a
to
0e2390e
Compare
ifed01
changed the title
ifed01
force-pushed
the
wip-bluestore-blob-release-fix
branch
from
0e2390e
to
75348da
Compare
|
retest this please |
|
Note: The above comment was to get jenkins to retest the build. |
|
Hi Igor, this branch has conflicts. Can you please fix? Thanks! |
ifed01
force-pushed
the
wip-bluestore-blob-release-fix
branch
from
75348da
to
e7162b8
Compare
|
@markhpc - rebased |
|
retest this please |
| int64_t id = get_new_id(); | ||
| Blob *b = new Blob(id, c); | ||
| b->get(); |
There was a problem hiding this comment.
If BlobRef is an intrusive_ptr there shouldn't be any explicit get/put calls anywhere (except the intrustive ptr hooks)...
There was a problem hiding this comment.
@liewegas. Additional ref_count increment is needed since Blobs are stored in intrusive::set blob_map. Blob has to persist while it's stored in this blob_map even if no corresponding BlobRefs are present.
Other options are either using std::set or intrusive::set but IMHO they both have some performance/memory footprint drawbacks. Have I missed something and there is better approach?
|
otherwise lgtm! |
Signed-off-by: Igor Fedotov <ifedotov@mirantis.com>
Signed-off-by: Igor Fedotov <ifedotov@mirantis.com>
Signed-off-by: Igor Fedotov <ifedotov@mirantis.com>
… onode::blob_map enumeration Signed-off-by: Igor Fedotov <ifedotov@mirantis.com>
ifed01
force-pushed
the
wip-bluestore-blob-release-fix
branch
from
e7162b8
to
6beade7
Compare
|
@liewegas squashed and rebased. |
This is an intermediate patch before introducing refcounting for Blob instances.
Signed-off-by: Igor Fedotov ifedotov@mirantis.com